Cost-Sensitive Radial Basis Function Neural Network Classifier for Software Defect Prediction

Table 7

Performance results of the proposed predictor on the five NASA datasets (non-cost-sensitive case).

NASA datasets	Sensitivity; (pd)	Specificity	False positive rate (FPR or pf)	Balance	Accuracy	Area under ROC (AUC)
CM1	81.92	80.96	29.71	75.41	82.57	0.90
JM1	79.85	82.31	36.22	70.69	77.03	0.87
KC1	85.67	87.95	20.24	82.46	84.96	0.92
KC2	87.96	86.24	17.93	84.73	88.65	0.95
PC1	77.84	89.33	30.23	73.49	86.29	0.93
Mean	82.65	85.36	26.87	77.36	83.90	0.9140